Download pdf - The impact of product market competition on earnings quality

The impact of product market competition onearnings quality

Peter Chenga, Paul Manb, Cheong H. Yic

aSchool of Accounting and Finance, The Hong Kong Polytechnic University,Hunghom, Hong Kong

bDepartment of Business Administration, Caritas Institute of Higher Education,Tseung Kwan O, Hong Kong

cDepartment of Accountancy, The City University of Hong Kong,Kowloon Tong, Hong Kong

Abstract

The objective of this paper is to examine the impact of product market competi-tion on earnings quality. Based on a sample from the US manufacturing sectorfor the period 1996–2005, we find consistent evidence showing a positive relationbetween product market competition and earnings quality. Additional tests alsoconfirm a positive relation between product market competition and the preci-sion of public and private information held by investors and analysts. We alsoprovide evidence that firms competing in concentrated and heterogeneous indus-tries are associated with a number of earnings attributes and information qualitynot shared by those competing in concentrated but homogeneous industries.These findings are consistent with the intuition that firms enjoying a monopolis-tic advantage tend to avoid the attention of their competitors and politicians bycreating a more opaque information environment.

Key words: Earnings quality; Industry concentration; Product market competi-tion

JEL classification: D80, L10, M40, M41

doi: 10.1111/j.1467-629X.2011.00457.x

The authors gratefully acknowledge the valuable comments from Steven Cahan (DeputyEditor), Charles Chen, Francis Kim, Donghui Wu and an anonymous reviewer.

Received 24 August 2010; accepted 25 October 2011 by Steve Cahan (Deputy Editor).

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Accounting and Finance

1. Introduction

The demand for financial reporting and disclosure arises from informationasymmetry and agency conflicts between managers and outside investors (Healyand Palepu, 2001). Yet, a manager’s motives for voluntary disclosure areaffected by economic, political and institutional determinants and settings suchas corporate control contests and shareholder litigation. Public disclosure ofinformation can affect the disclosing firm negatively if market participants makestrategic use of the information to their advantage, or if such disclosure rendersthe firm a political target. In the presence of such proprietary and political costs,firms may opt to protect their competitive advantage by exercising control overwhat information is voluntarily disclosed. However, empirical evidence in theliterature mainly centres on the quantity aspect of disclosure such as thefrequency and horizons of forecasts (Ali et al., 2010), number of segments to bereported (Harris, 1998; Botosan and Stanford, 2005) and the pervasiveness offorecasting in the industry (Li, 2010). As far as disclosure quality is concerned,prior studies have focused on the choice of income-deflating accounting policies(Hagerman and Zmijewski, 1979) and forecast accuracy (Ali et al., 2010;Li, 2010).The objective of this paper is to add to this literature by examining the impact

of product market competition, as measured by industry concentration, on thequality of accounting information. Drawing data from the US manufacturingindustry and using various earnings attributes as proxies for accounting infor-mation quality, we find a consistent and significantly negative relation betweenindustry concentration and various earnings attributes: accruals quality, persis-tence, predictability, smoothness, value relevance, timeliness and conservatism.These findings are consistent with the intuition that firms in concentrated indus-tries tend to protect their competitive position by releasing lower quality earn-ings information. We provide further evidence that the earnings of firmsoperating in concentrated and heterogeneous industries are associated with alower level of accruals quality, predictability, value relevance and timelinessthan are those of firms operating in concentrated and homogeneous industries.Again, this supports the notion that firms facing less competition tend to createa more opaque information environment to protect their competitive advantage.Additional tests suggest that the above findings apply not only to earnings qual-ity but also to the precision of private and public information held by investorsand analysts.This paper contributes to the literature in three ways. To begin, this is the

first study to provide evidence on a positive relation between product marketcompetition and numerous attributes of earnings quality. This finding sup-ports the notion that firms facing less competition tend to create an opaqueinformation environment to protect their competitive advantage over rivalsand avoid public and political sanctions. Second, prior research (Hagermanand Zmijewski, 1979; Ali et al., 2010; Li, 2010) on the impact of competition

2 P. Cheng et al./Accounting and Finance


on disclosure quality is limited to the examination of accounting policy choiceand the accuracy of analysts’ forecasts. This paper extends this stream of lit-erature by testing a different dimension of the information environment: theprecision of public and private information held by investors and analysts.Finally, to the best of our knowledge, this is the first paper to incorporateindustry homogeneity into the analysis to buttress the relation between mar-ket concentration and information quality.The remainder of this paper proceeds as follows. In Section 2, we first review

the literature on selected motives that constrain voluntary disclosure and theirrelations with product market competition before proceeding to develop our test-able hypotheses. In Section 3, we introduce our research design and describe ourvariable definitions and measures. The empirical findings are discussed in detailin Section 4, and we present the results of additional and sensitivity tests inSection 5. We draw our conclusions in Section 6.

2. Prior literature and hypothesis development

2.1 Prior literature

Our study of the relation between product market competition and earningsquality begins with the structure-conduct-performance (SCP) paradigm,1 whichhas been widely applied in the field of industrial organization research. Thisapproach assumes that the smaller the number of firms in an industry (asreflected in the market shares of individual firms within that industry), thegreater the likelihood that market power will be abused through collusive behav-iour and higher profitability among established firms.2 While the SCP approachassumes that competition increases as the number of firms within an industryrises, economic theory suggests that the degree of competition within a concen-trated industry also depends on the level of product homogeneity between com-petitors. In the typical Bertrand oligopoly model, firms offer homogeneousproducts and face the same downward sloping demand curve. Competitionbetween firms is intense, and in the Nash equilibrium, they charge a price equalto marginal cost and earn zero profit. On the other hand, in monopolistic com-petition, firms offer differentiated products, and each firm enjoys a certain degree

1 This paradigm was developed by Bain (1968). See Weiss (1979) for an overview of theparadigm.

2 A large number of studies have tested this assumption empirically. Using data from USmanufacturing industries, Bain (1951) finds a positive relation between firm return onequity and industry concentration level. See Weiss (1974) for a review of other studieswhich also report a significant and positive relationship between profits and industryconcentration.

P. Cheng et al./Accounting and Finance 3


of monopolistic advantage in the market for its particular product variant.3

Similarly, industries made up of firms similar in terms of strategy, structure ortechnological level4 will exhibit a higher level of competition. Consequently,other things being equal, a higher (lower) level of industry concentration and alower (higher) level of industry homogeneity5 are associated with a higher (lower)level of monopolistic advantage.While firms from concentrated industries are associated with higher profitabil-

ity, the literature also indicates that such firms face higher proprietary disclosureand political costs. Theoretical models developed by Grossman and Hart (1980),Grossman (1981) and Milgrom (1981) show that one of the conditions for firmsvoluntarily disclosing all their private information is that disclosures have to becostless. Yet, an extensive body of research shows that voluntary disclosure isnot without cost and that managers’ decisions on the optimum level of disclosureare affected by economic, institutional and political factors. The proprietary-costmotive postulates that managers’ voluntary disclosures to investors may damagethe competitive position of the firm by releasing sensitive information to compet-itors (Verrecchia, 1983). Managers’ decisions on the appropriate level of disclo-sure involve a trade-off between the benefits of informing the capital marketabout firm value and the costs of aiding the firm’s rivals. Darrough and Strough-ton (1990) theoretically demonstrate that firms in less competitive industries haveless of an incentive to make informative disclosures for fear of attracting compe-tition. As noted earlier, the SCP model assumes that industry concentrationweakens competition by fostering collusive behaviour among firms. In addition,firms from concentrated industries that sell differentiated products enjoy a higherlevel of monopolistic advantage. Because competition drives away abnormalprofits,6 firms in concentrated and heterogeneous industries are associated withhigher proprietary costs.In addition to incurring higher proprietary disclosure costs, firms from concen-

trated and heterogeneous industries may also be subject to higher political costs.Watts and Zimmerman (1978) predict that managers confronted with the possi-bility of politically imposed wealth transfers will choose accounting strategiesthat reduce the likelihood or size of the transfer. Antitrust laws in the United

3 See, for example, Jehle and Reny (2000) on the discussion of the various models underimperfect competition.

4 Porter (1979) refers to these firms as ‘strategic groups’; Aghion et al. (2005) label themthe ‘neck-and-neck sector’.

5 Industry homogeneity refers to the extent to which an industry is made up of firms simi-lar in nature, technology or products.

6 The competitive environment hypothesis states that competition will eliminate all abnor-mal profits in the long run (Schumpeter, 1934). Extensive research on trends in companyprofits over time is supportive of this mean-reversion phenomenon. See, for example,Roberts and Dowling (2002).



States prohibit monopolies or attempt to create a monopoly in any unregulatedline of business.7 High accounting rates of return are regarded as ‘excessive’ andindicative of monopolistic power on the part of the firm. The political-costmotive suggests that firms in concentrated industries are associated with higherpolitical costs and will have more of an incentive to employ earnings manage-ment techniques to reduce reported profits. Adopting a longitudinal approachthat examines likely changes in firms’ political costs over time, Cahan (1992)finds that managers adjust their discretionary accruals in response to monopoly-related antitrust investigations, thus supporting the political-cost motive.Both the proprietary-cost and political-cost motives suggest that managers will

avoid full disclosure to protect their respective firm values. To do so, managersmay choose to control the amount (quantity) and/or quality of information tobe disclosed. Harris (1998) and Botosan and Stanford (2005) provide empiricalevidence demonstrating that firms in concentrated industries control the quantityof accounting information to be revealed by hiding information on profitablesegments. Ali et al. (2010) also find that oligopolistic industries provide manage-ment earnings forecasts less frequently and that their forecasts have shorter hori-zons. Using industry pervasiveness of forecasting as a proxy for disclosurequantity, Li (2010) shows that competition from existing rivals reduces the quan-tity of information disclosed.To avoid revealing precise information on financial performance to their rivals

and the general public, managers may also choose to create an opaque informa-tion environment by controlling the quality – rather than the quantity – of infor-mation to be released. Using probit analysis, Hagerman and Zmijewski (1979)find that firms from concentrated industries are more likely to adopt accountingpolicies that minimize reported income. Both Ali et al. (2010) and Li (2010)examine the impact of competition on the quality of information released byfirms by using properties of analysts’ forecasts as a proxy for quality of informa-tion. In this paper, we use earnings quality as a proxy for the quality of account-ing information and conjecture that firms from less competitive industries areassociated with a lower level of accounting information quality.

2.2 Hypothesis development

The foregoing discussion suggests that firms from concentrated and hetero-geneous industries are associated with higher disclosure costs and are morelikely to offer financial information of lower quality to avoid the attention of

7 The Sherman Antitrust Act (1890) makes it a criminal offence to monopolize any partof interstate commerce. An unlawful monopoly exists when only one firm controls themarket for a product or service and has obtained that market power by suppressing com-petition through anticompetitive conduct rather than because its product or serviceis superior to others. In addition, the Clayton Act (1910) prohibits mergers or acquisitionslikely to lessen competition.



rivals and sanctions from the public. Following prior research in accounting,8

we employ earnings quality as a summary indicator of and proxy for finan-cial reporting quality. We employ seven attributes (accruals quality, earningspersistence, earnings predictability, earnings smoothness, value relevance,earnings timeliness and earnings conservatism) as proxies for earnings qualityto capture differences in underlying assumptions about the function of earn-ings.9

Accruals quality captures the extent to which current accruals are mappedinto operating cash flow realizations and changes in revenues and fixed assets.A poor match signifies low accruals quality, that is, estimation errors and subse-quent corrections reduce the beneficial role of accruals, which in turn affectsmapping between earnings and cash flows. Prior studies (Jones, 1991; Warfieldet al., 1995; Subramanyam, 1996; Lim and Matolcsy, 1999) provide evidence onthe opportunistic use of discretionary accruals by management to window dressand mislead users of financial statements. Their findings suggest that managerialintent affects the incidence and magnitude of accrual estimation errors. Basedon both the proprietary-cost and political-cost motives, we conjecture that firmsfrom less competitive industries will exhibit larger accrual estimation errors andpoorer mapping between earnings and cash flows, as their managers are morelikely to make use of discretionary accruals to achieve earnings managementgoals.Earnings persistence captures the effect of earnings innovations, that is, new

information in earnings in relation to expected future earnings (Miller and Rock,1985; Kormendi and Lipe, 1987). Earnings that are more persistent or sustain-able are indicative of a firm’s long-run earning ability (Penman, 2001). Similarly,earnings predictability captures the ability to use past earnings to predict futureearnings (Lipe, 1990). Persistent and predictable earnings among firms in lesscompetitive industries are more likely to encourage potential competitors toenter the market (Qualls, 1974; Scherer, 1980; Jacobsen, 1988) and attract sanc-tions from political bodies and the general public (Han and Wang, 1998).We therefore conjecture that firms from less competitive industries will exhibit

8 See, for example, Core et al. (2008), Cohen (2008), Francis et al. (2004, 2005), Franciset al. (2006) and Gaio and Raposo (2011).

9 Accounting-based earnings attributes include accruals quality, earnings persistence,earnings predictability, and earnings smoothness. These four attributes take either cash orearnings (or other measures that can be derived from these measures, such as accruals)as the reference construct, and are estimated using accounting data. Market-based attri-butes are value relevance, earnings timeliness, and earnings conservatism. These threeattributes take returns or prices as the reference construct and rely on both accountingand returns data for their estimation. Accounting-based earnings quality measures assumethat the function of earnings is to allocate cash flows to reporting periods via accruals,while market-based earnings quality measures assume that the function of earnings is toreflect economic income as represented by stock returns.



less persistent and/or less predictable earnings as their managers are more likelyto adopt strategies that conceal their earnings persistency and predictability.Earnings smoothness reflects the idea that managers use their private informa-

tion about future income to smooth out transitory fluctuations and noise,thereby achieving a more representative and useful earnings number for investorsto assess the value of the firm (Francis et al., 2006). To the extent that firm man-agers from concentrated and heterogeneous industries would like to reduce theinformativeness of earnings to avoid the attention of competitors and the public,we conjecture that earnings among such firms are less smooth. In addition,by reducing the variance of the firm’s underlying economic earnings, Truemanand Titman (1988) argue that income smoothing reduces the firm’s cost of bor-rowing. Empirically, Francis et al. (2004, 2005) find that capital marketparticipants reward smoother earnings streams with reduced costs of equity anddebt capital. Because more intense competition increases the likelihood of firmliquidation (Hou and Robinson, 2006), firms from less concentrated and homo-geneous industries have a stronger demand for efficient contracting, and theirmanagers have a stronger incentive to engage in earnings smoothing activities togain access to funds at a lower cost.Value relevance measures the ability to use one or more accounting numbers

to explain variations in stock returns. Earnings with greater explanatory powerare viewed as being of higher quality. In a setting in which managers have bothincentives and opportunities to manage earnings, Marquardt and Wiedman(2004) find a decrease in the value relevance of earnings when earnings manage-ment is present. To the extent that managers from concentrated industries exhi-bit opportunistic behaviours to avoid proprietary and political costs, we predictthat the earnings of their firms will be less value relevant. Earnings timelinesscaptures the intrinsic lead/lag relation between earnings and returns (Gelb andZarowin, 2002). Firms that report earnings on a more (less) timely basis shouldhave a stronger (weaker) relation between returns and current earnings. GarciaLara et al. (2005) show that the asymmetric timeliness of earnings is affected bythe extent and pervasiveness of earnings management. Because managers fromconcentrated and heterogeneous industries will be more inclined to take upaccrual choices that lead to a lower reported earnings number, we predict thatthe contemporaneous association between returns and current earnings will beweaker for such firms.The final attribute of conservatism reflects the differing extent to which

accounting earnings reflect economic losses as opposed to economic gains (Basu,1997). In other words, earnings reflect bad news more quickly than they do goodnews. Watts (2003) argues that accounting conservatism is a desirable earningsattribute because conservative reporting constrains opportunistic behaviouramong managers and offsets managerial biases with its asymmetrical verifiabilityrequirement. Dhaliwal et al. (2008) provide evidence that the degree of account-ing conservatism increases with the intensity of competition, supporting the ideathat competition increases the demand for more efficient contracting and the risk



of liquidation. Their findings are also consistent with the argument that competi-tion improves the flow of firm-specific information by enabling easier evaluationof a firm’s performance vis-a-vis its peers (Holmstrom, 1982; Nalebuff and Sti-glitz, 1983). Consequently, managers from intensely competitive industries havea limited ability to conceal bad news and are therefore induced to recognize eco-nomic losses in a timely manner.The above-mentioned discussion suggests that lower level of product market

competition is likely to lead to lower earnings quality, reflected in lower accrualsquality, lower earnings persistence, less predictable earnings, higher incomesmoothing, lower value relevance of earnings, less timely earnings and a lowerlevel of accounting conservatism. Moreover, the positive relationship betweenproduct market competition and earnings quality is expected to be more pro-nounced in less homogeneous industries. While we examine these relationshipsusing each of the seven earnings attributes, we summarize our expectations usingthe following hypotheses:

Hypothesis 1a: The higher the level of industry concentration, ceteris paribus,the lower the quality of firms’ earnings.

Hypothesis 1b: The higher the level of industry concentration and the higher thelevel of industry homogeneity, ceteris paribus, the higher the quality of firms’earnings.

3. Research design and variable measurement

3.1 Research design

Tests of parts (a) and (b) of our hypothesis are based on the following empiri-cal equations, respectively:

QUALITY ¼ b0 þ b1CONCþ Controlsþ Yr-Dummies

þ Ind-Dummiesþ e ð1Þ

QUALITY ¼ b0 þ b1CONCþ b2HOMOþ b3CONC �HOMO

þ Controlsþ Yr-Dummiesþ Ind-Dummiesþ e ð2Þ

The first equation examines the relation between industry concentrationand the quality of accounting information, while the second equation furtherexamines the incremental effect of industry concentration on the quality ofaccounting information when firms are competing in homogeneous industries.We use earnings quality to measure the quality of accounting information(QUALITY) as represented by various earnings attributes discussed by



Francis et al. (2004, 2005) and Francis et al. (2006). CONC is the proxy forindustry concentration, and HOMO is a dummy variable that takes the valueof one when the industry homogeneity measure is above the sample mean.In equation (1), the coefficient b1 measures the effect of industry concentra-tion on earnings quality without regard to industry structure. In equation (2),the coefficient b1 captures the relation between industry concentration andearnings quality for firms in heterogeneous industries, while the coefficient b3measures the joint incremental effect of industry concentration on earningsquality in homogeneous industries.As explained by Dechow and Dichev (2002) and Francis et al. (2004, 2005),

earnings attributes are jointly determined by intrinsic (innate) factors and man-agement’s (discretionary) reporting and implementation decisions. We thereforefollow Dechow and Dichev (2002) and Francis et al. (2004, 2005) by including inour above-mentioned equations the following control variables: firm size (SIZE),cash flow variability (r(CFO)), sales variability (r(SALES)), length of operatingcycle (OP-CYCLE), incidence of negative earnings realizations (NEG-EARN),leverage (LEVERAGE), market-to-book ratio (MB), intangibles intensity (INT-INTENSITY), absence of intangibles (INT-DUMMY) and capital intensity(CAP-INTENSITY). Industry and year dummy variables are included to controlfor industry and year effects.

3.2 Variable measurement

3.2.1 Earnings attributes

We follow Francis et al. (2004, 2005) for the measurement and estimation ofthe seven earnings attributes and add C-SCORE as an additional proxy to theBasu (1997) measure for conservatism.10 Briefly, accruals quality (AQ) is the neg-ative of the standard deviation of the firm’s residuals from a regression of currentaccruals on lagged, current and future cash flows from operations, change in rev-enues and gross value of property, plant and equipment; earnings persistence(PERSISTENCE) is the slope coefficient of the firm’s AR1 model of annualearnings; earnings predictability (PREDICTABILITY) is the negative of thesquare root of the error variance from the firm’s AR1 model of annual earnings;earnings smoothness (SMOOTHNESS) is the negative of the ratio of the firm’sstandard deviation of earnings before extraordinary items to the standard devia-tion of cash flows from operations; value relevance (RELEVANCE) is the

10 As discussed by LaFond and Watts (2008) and Khan and Watts (2009), the traditionalBasu measure of conservatism obscures the timing of changes in the conservatism offinancial reports issued by individual firms by assuming the firm’s operating characteris-tics to be stationary. However, changes affecting a firm’s financial reporting conservatismare likely to be both time- and firm-specific. We thus follow Khan and Watts (2009) andemploy C-SCORE as our second measure of conservatism.



adjusted R2 from a regression of returns on the level of and change in the firm’sannual earnings before extraordinary items; earnings timeliness (TIMELINESS)is the adjusted R2 from a reverse regression of the firm’s annual earnings beforeextraordinary items on variables capturing positive and negative returns; earn-ings conservatism (CONSERVATISM) is the ratio of the coefficient on the firm’sbad news (negative returns) to good news (positive returns) in the reverse regres-sion; C-SCORE is the incremental timeliness of the firm’s bad news each year.We adjust the signs of these attributes such that the results can be interpreted aslarger values of the attribute indicating more favourable outcomes, that is, higherearnings quality.

3.2.2 Industry concentration

Industry concentration is measured using the Herfindahl index, calculated asthe sum of the squared market shares of all firms in the market. In calculatingthe index, a number of studies use data from Compustat, which covers only pub-lic firms. A growing number of studies,11 however, have argued that such Com-pustat-based measures are subject to sampling bias because of the exclusion ofprivate firms from the calculations. As pointed out by Hay and Morris (1991)and Ali et al. (2009), industry concentration measures calculated using Compu-stat data may provide an inaccurate picture of the actual degree of concentrationin industries, particularly for those in which private firms account for a non-neg-ligible percentage of industry sales.In view of this, we use data on the Herfindahl index collected from Census

of Manufactures publications provided by the US Census Bureau. The Herfin-dahl index is calculated for 6-digit NAICS industries within the manufactur-ing sector.12 Because the Census of Manufactures is published only duringyears when a US census takes place, we follow several prior studies13 anduse the census data for a given year as a proxy for industry concentrationnot only for that year but also for the 5-year window surrounding the yearin which the census takes place. In other words, census data reported in 1997(2002) are used for the sample period of 1995 to 1999 (2000–2005). Thisapproach is taken based on the rationale that structural characteristics of anindustry, such as concentration, change slowly over time (Caves, 1980; Lip-czvnski and Wilson, 2001). Using data from the 1997 and 2002 US censuses,our variable for the census-based Herfindahl index, CONC, covers the sampleperiod of 1996 to 2005.

11 See, for example, MacKay and Phillips (2005), Campello (2006), Akdogu and Mackay(2008), and Ali et al. (2009).

12 The sample covers firms with 6-digit NAICS codes between 311111 and 339999.

13 See, for example, Aggarwal and Samwick (1999), MacKay and Phillips (2005),Campello (2006), Haushalter et al. (2007), and Ali et al. (2009).



3.2.3 Industry homogeneity

Following Parrino (1997), our proxy for industry homogeneity, HOMO-GENEITY, measures the correlation among firms’ common stock returns withinthe two-digit SIC industries.14 A measure based on changes in stock prices is anatural choice for an industry homogeneity proxy because a firm’s stock pricereflects the present value of its residual cash flow. If the firms in an industryemploy similar production technologies and compete in similar product markets,news concerning changes in factors such as economic conditions or technologicalshocks will tend to affect their cash flows, and therefore their stock prices, in asimilar manner. The higher the value of HOMOGENEITY, the more homo-geneous the firms within the industry. To investigate the effect of industry homo-geneity on the relation between industry concentration and earnings quality, weemploy a dummy variable HOMO (which takes the value of 1 when HOMO-GENEITY is above the sample mean, and 0 otherwise) and interact it withCONC to capture the incremental effect.

3.2.4 Control variables and innate factors

Following Francis et al. (2004, 2005), SIZE is the log of the firm’s total assets;r(CFO) is the standard deviation of the firm’s rolling 10-year cash flows fromoperations; r(SALES) is the standard deviation of the firm’s rolling 10-yearsales revenues; OP-CYCLE is the log of the sum of the firm’s days of accountsreceivable and days of inventory; NEG-EARN is the firm’s proportion of lossesover the prior 10 years; LEVERAGE is the total of the firm’s long- and short-term debt scaled by its market value of equity; MB is the firm’s market value ofequity divided by its book value of equity; INT-INTENSITY is the sum of thefirm’s reported R&D and advertising expenses as a proportion of its sales reve-nues; INT-DUMMY is a dummy variable that takes the value of 1 when thefirm’s INT-INTENSITY equals zero and 0 otherwise; and CAP-INTENSITY isthe ratio of the firm’s net book value of property, plant and equipment to totalassets.Based on our suggested hypotheses and the measurement of the various vari-

ables, we expect a negative relation between CONC and earnings quality, that is,a negative value for b1 under both equations (1) and (2) mentioned earlier.Moreover, because we expect a higher level of competition when firms arecompeting in homogeneous industries, we expect a positive value for b3 underequation (2). As for the control variables, we expect a positive coefficient

14 HOMOGENEITY is the average of the partial correlation coefficient for all firmswithin a 2-digit SIC industry from the regression Rjt = b0 + b1Rit + b2Rmt + ejt, whereRjt = stock return for firm j in industry i for month t; Rit = equally weighted stockreturn for industry i in month t; and Rmt = equally weighted stock return for the marketin month t.



for r(CFO), r(SALES), OP-CYCLE, NEG-EARN, LEVERAGE andINT-INTENSITY and a negative coefficient for SIZE, MB, INT-DUMMY andCAP-INTENSITY.

4. Empirical findings

4.1 Sample and descriptive statistics

Stock returns and financial statement data are collected from the Center forResearch in Security Prices (CRSP) and Compustat databases, respectively. Ana-lysts’ earnings forecast data are collected from the I/B/E/S database. Becausemost of the earnings attributes are calculated over rolling firm-specific 10-yearwindows, a firm is included in the year t sample only if data are available fromyear t-9 through to year t. Consistent with prior studies, we winsorize theextreme values of the distribution at the 1 and 99 percentiles, and we furtherrequire that data on all eight proxies are available for each firm-year. A total of4989 firm-year observations from 976 distinct firms covering the years 1996 to2005 satisfy these requirements.Table 1 presents the distributions of firms by year and the 2-digit SIC codes in

our final sample. 1997 is the year with the highest number of firms (566), while2005 has the lowest (411). We further note that firm-year observations from thebiggest three industries (SIC codes 35, 36 and 28) alone represent over 58 percent of the total sample. A financial summary of our sample is presented inTable 2. The mean (median) market value of $5292 ($616) million and the mean(median) total assets of $3036 ($505) million indicate that our sample consists ofrelatively large firms. Moreover, these huge variations between mean and medianvalues, together with the large standard deviations ($19,897 million in marketvalue and $9720 million in asset value), indicate that the size of firms in oursample is highly skewed. To avoid having our test results dominated by firm sizeor industry grouping, we include several robustness tests in our regression resultsin Section 5. Descriptive statistics on ROA, MB and LEVERAGE are consistentwith the characteristics of firms from the manufacturing sector.15

The same table shows that the mean and median values of CONC are 0.059and 0.048, respectively, suggesting that the sample represents rather competitive

15 Table 2 reveals that most of the firms have low ROA with a mean value of 0.028 andhigh MB with a mean value of 3.466. These results respectively reflect the capital- andR&D-intensive nature of firms in the manufacturing sector. As pointed out by Chan et al.(2001), as a result of the expensing convention for R&D, some yardsticks commonly usedby investors, such as price-earnings ratios and market-to-book ratios, may be misstated.In particular, many R&D-intensive companies may appear to be priced at unjustifiablyhigh multiples. LEVERAGE is low with a mean value of 0.288, and is consistent with thefinding that debt is not a favored form of finance for R&D-intensive manufacturing firms(Hall, 1992).



Table 1

Sample description, 1996–2005

Main sample Forecast sample

Panel A: Number of firm-years

Firm-years for 1995–2006

Less: Firm-years used as lead/lag values for variables

12 098

(1346)

12 098

(1346)

Firm-years for 1996–2005

Less: Missing data in earnings attributes/IBES

10 752

(5763)

10 752

(5012)

Firm-years in final sample 4989 5740

Panel B: Number of firms by year

1996 556 729

1997 566 703

1998 523 677

1999 502 612

2000 477 554

2001 467 504

2002 477 477

2003 501 484

2004 509 500

2005 411 500

Total 4989 5740

Distinct 976 929

Panel C: Number of firms by 2-digit SIC code

20 – Food and kindred products 163 217

22 – Textile mill products 9 31

23 – Apparel and other textile products 88 108

24 – Lumber and wood products – 30

25 – Furniture and fixtures 46 56

26 – Paper and allied products 74 108

27 – Printing and publishing 37 49

28 – Chemical and allied products 714 963

29 – Petroleum and coal products 59 122

30 – Rubber and miscellaneous plastic products 142 118

31 – Leather and leather products 71 71

32 – Stone, clay and glass products – 29

33 – Primary metal industries 157 234

34 – Fabricated metal products 170 160

35 – Industrial machinery and equipment 820 970

36 – Electronic and other electric equipment 1075 1135

37 – Transportation equipment 237 291

38 – Instruments and related products 1008 933

39 – Miscellaneous manufacturing industries 119 115

Total 4989 5740



industries. HOMOGENEITY has a sample mean of 0.190, indicating that firmswithin our sample are mainly competing within a heterogeneous environment.16

Turning to the earnings attributes, their absolute sample mean values are asollows: AQ = 0.044, PERSISTENCE = 0.310, PREDICTABILITY = 0.696,SMOOTHNESS = 0.807, RELEVANCE = 0.438, TIMELINESS = 0.488and CONSERVATISM = 1.38. With the exception of that for

Table 2

Sample description and descriptive statistics on financial & key variables, 1996–2005

Mean SD 10% 25% Median 75% 90%

Financial variables

Mkt. value of equity ($mils) 5292 19,897 76 186 616 2419 9113

Assets ($mils) 3036 9720 59 150 505 1843 6694

Sales ($mils) 3001 10,991 46 144 516 1813 6594

ROA 0.028 0.154 )0.094 0.012 0.055 0.096 0.136

MB 3.466 3.635 1.118 1.624 2.466 3.855 6.472

Leverage 0.288 0.562 0 0.015 0.126 0.335 0.684

Earnings per share 1.006 1.702 )0.740 0.150 0.950 1.800 2.850

CONC 0.059 0.046 0.012 0.024 0.048 0.081 0.125

HOMOGENEITY 0.190 0.038 0.159 0.160 0.186 0.202 0.234

Earnings attributes

AQ )0.044 0.035 )0.083 )0.054 )0.034 )0.023 )0.015PERSISTENCE 0.310 0.406 )0.174 0.070 0.320 0.554 0.783

PREDICTABILITY )0.696 0.496 )1.376 )0.928 )0.573 )0.327 )0.191SMOOTHNESS )0.807 0.401 )1.318 )1.051 )0.760 )0.494 )0.340RELEVANCE 0.438 0.252 0.093 0.228 0.431 0.640 0.790

TIMELINESS 0.488 0.252 0.134 0.281 0.489 0.693 0.830

CONSERVATISM 1.380 413.117 )9.189 )1.646 0.980 3.380 11.635

C-SCORE 0.116 0.101 )0.006 0.053 0.108 0.175 0.241

Innate factors

SIZE 5.469 1.850 3.110 4.006 5.404 6.775 7.961

r(CFO) 0.082 0.075 0.026 0.041 0.064 0.101 0.153

r(SALES) 0.104 0.083 0.027 0.045 0.081 0.137 0.213

OP-CYCLE 5.030 0.457 4.480 4.761 5.032 5.315 5.561

NEG-EARN 0.202 0.277 0 0 0 0.400 0.600

INT-INTENSITY 0.116 0.816 0 0.010 0.038 0.097 0.189

INT-DUMMY 0.159 0.366 – – – – –

CAP-INTENSITY 0.233 0.143 0.073 0.128 0.204 0.312 0.429

16 Based on a sample period of 1970–1988, Parrino (1997) computes the homogeneitymeasure for each 2-digit SIC industry and finds that their values range from 0.1522 to0.5276, with a mean value of 0.2974 and a median of 0.2823. As the structural characteris-tics of an industry change slowly over time, the sample mean of our homogeneity measureof 0.190 indicates that most of our sample firms come from heterogeneous industries.



CONSERVATISM, these values are similar to those reported by Francis et al.(2004, 2005). As for CONSERVATISM, the discrepancy in the sample meanvalue could be due to the fact that our sample is mostly made up of large manu-facturing firms. Its large sample standard deviation of 413 is driven by observa-tions with small values of the denominator of this variable, which is the ratio ofthe coefficient on bad news to the coefficient on good news. As for C-SCORE,the sample mean and median of 0.116 and 0.108 are similar to those reported(0.105 and 0.097) by Khan and Watts (2009). The summary statistics for thecontrol variables are similar to those reported by Francis et al. (2004, 2005). Uni-variate tests (untabulated) of the earnings attributes indicate that they are not

Table 2 (continued)

Mean SD 10% 25% Median 75% 90%

Information qualityffiffiffi

hp

22.078 54.479 0.173 2.685 10.025 24.159 46.046ffiffi

sp

21.791 29.569 0 2.107 9.875 29.680 60.443ffiffiffiffiffiffiffiffiffiffiffi

hþ sp

37.257 58.450 2.994 8.955 22.018 47.958 84.984

Sample description and variable definitions: The sample contains 4989 firm-year observations over

t = 1996–2005 and covers firms with 6-digit NAICS codes between 311111 and 339999. ROA =

return on assets; MB = market value of equity divided by book value of equity; Leverage = total

of long- and short-term debt scaled by market value of equity; CONC is the Herfindahl index for

6-digit NAICS industries as reported by the Census of Manufactures publications. HOMOGENEITY

is the average of the partial correlation coefficient for all firms within a 2-digit SIC industry from the

following regression: Rjt = b0 + b1Rit + b2Rmt + ejt, where Rjt = stock return for firm j in

industry i for month t; Rit = equally weighted stock return for industry i in month t; and Rmt =

equally weighted stock return for market in month t; AQ = negative of the standard deviation of

firm’s residuals from a regression of current accruals on lagged, current and future cash flows from

operations, change in revenues and gross value of property, plant and equipment; PERSIS-

TENCE = slope coefficient of firm’s AR1 model of annual earnings; PREDICTABILITY = nega-

tive of the square root of the error variance from firm’s AR1 model of annual earnings;

SMOOTHNESS = negative of the ratio of firm’s standard deviation of earnings before extraordi-

nary items to the standard deviation of cash flows from operations; RELEVANCE = adjusted

R2 from a regression of 15-month returns on the level and change in firm’s annual earnings before

extraordinary items; TIMELINESS = adjusted R2 from a reverse regression of firm’s annual earn-

ings before extraordinary items on variables capturing positive and negative 15-month returns;

CONSERVATISM = ratio of the coefficient on firm’s bad news (negative returns) to coefficient on

good news (positive returns) in the reverse regression; C-SCORE = incremental timeliness of firm’s

bad news each year (refer to Khan and Watts (2009) on estimation procedures); SIZE = log of

firm’s total assets; r(CFO) = standard deviation of firm’s rolling 10-year cash flows from opera-

tions; r(SALES) = standard deviation of firm’s rolling 10-year sales revenues; OP-CYCLE = log

of the sum of firm’s days of accounts receivable and days of inventory; NEG-EARN = proportion

of firm’s losses over the prior 10 years; INT-INTENSITY = sum of firm’s reported R&D and

advertising expenses as a proportion of its sales revenues; INT-DUMMY = 1 when firm’s

INT-INTENSITY equals zero, and 0 otherwise; CAP-INTENSITY = ratio of firm’s net book value

of property, plant and equipment to total assets;ffiffiffi

hp

= precision of public information;ffiffi

sp

= preci-

sion of private information;ffiffiffi

hpþ s = precision of total information (refer to Barron et al. (2002)

for measurement and estimation of h and s).



normally distributed. To avoid distorting our test results, our regressions arebased on the decile ranks of the earnings attributes, while actual values are usedfor the test and control variables.

4.2 Correlations

Pearson correlations among the earnings attributes and their correlations withindustry concentration are reported in Table 3. There exist some significant cor-relations between the earnings attributes, for example, AQ and SMOOTHNESS(0.25, p < 0.0001), RELEVANCE and TIMELINESS (0.67, p < 0.0001). Toinvestigate the degree of overlap among these attributes, we follow Francis et al.(2004, 2005) and calculate the auxiliary R2 statistics for each attribute. The auxil-iary R2 for attribute k is the R2 obtained from regressing attribute k on the othersix attributes.17 As indicated in the far right column of Table 3, the auxiliary R2

statistics for the various attributes are generally low, with the highest being 0.45for TIMELINESS, indicating that the degree of overlap among the attributes isnot significant. We therefore conclude that the various earnings attributes aredistinct constructs and that each attribute captures different aspects of earningsquality. Turning to correlations between industry concentration and the earningsattributes, significant negative correlations exist between CONC and almost allearnings attributes, for example, AQ = )0.064, SMOOTHNESS = )0.129 andC-SCORE = )0.05. These negative associations indicate that firms in concen-trated industries are associated with lower earnings quality and provide prelimin-ary support for our predictions.

4.3 Regression results

The results from the estimation of equation (1) are reported in Table 4. Theresults indicate that the coefficients on the control variables are highly significantand display the predicted sign. These results are generally consistent with thosereported by Dechow and Dichev (2002) and Francis et al. (2004, 2005).Turning to the main variable of interest, CONC, its coefficient estimates are

quite consistent across the different earnings attributes. As shown in Table 4, theestimated coefficients of CONC across the earnings attributes other thanC-SCORE are all negative and significant, for example, AQ = )2.236 (t-value)3.09), SMOOTHNESS = )6.042 (t-value )7.69) and RELEVANCE = )2.799(t-value )3.31). The above findings are consistent with the notion that firms inconcentrated industries are associated with a lower level of earnings quality andare supportive of our Hypothesis 1(a). Table 5 reports the results from the

17 As explained by Francis et al. (2004), the auxiliary R2 is analogous to a multivariatecorrelation between each attribute and all other attributes – the square root of the auxil-iary R2 is the Pearson multivariate correlation.



Table

3

Pearsoncorrelationsam

ongearnings

attributesan

dindustry

concentration(N

=4989)

AQ

PERSIS-

TENCE

PREDIC

T-

ABIL

ITY

SMOOTH-

NESS

RELEV-

ANCE

TIM

E-

LIN

ESS

CONSER-

VATISM

C-SCORE

CONC

Aux

iliary

R2

AQ

1.0000

0.0209

0.1407

0.0181

0.2024

0.2506

<0.0001

0.0439

0.0019

0.0742

<0.0001

)0.0065

0.6452

)0.1887

<0.0001

)0.0635

<0.0001

0.095

PERSISTENCE

1.0000

0.1168

<0.0001

)0.0096

0.4986

0.0341

0.0159

0.0327

0.0209

)0.0046

0.7481

0.0212

0.1352

)0.0214

0.1310

0.017

PREDIC

TABIL

ITY

1.0000

0.2064

<0.0001

0.0984

<0.0001

0.1077

<0.0001

)0.0057

0.6861

0.2209

<0.0001

)0.0791

<0.0001

0.115

SMOOTHNESS

1.0000

0.1380

<0.0001

0.1723

<0.0001

0.0168

0.2351

)0.0663

<0.0001

)0.1289

<0.0001

0.129

RELEVANCE

1.0000

0.6652

<0.0001

)0.0107

0.4510

0.0413

0.0035

)0.0720

<0.0001

0.444

TIM

ELIN

ESS

1.0000

0.0038

0.7899

0.0314

0.0267

)0.0616

<0.0001

0.451

CONSERVATISM

1.0000

)0.0038

0.7914

)0.0079

0.5775

0.001

C-SCORE

1.0000

)0.0471

0.0009

0.092

Refer

toTab

le2forvariab

ledefinitions.Significance

levelsareshownin

italics.Theau

xiliary

R2istheR2from

regressingattribute

kontheother

sixattri-

butes.



Table

4

Resultsofcross-sectional

regressionofdecileranksofearnings

attributesonindustry

concentration

Dependentvariable(N

=4989)

AQ

PERSIS-

TENCE

PREDIC

T-

ABIL

ITY

SMOOTH-

NESS

RELE-

VANCE

TIM

EL-

INESS

CONSER-

VATISM

C-SCORE

Intercept

5.925

(11.85)***

4.019

(6.81)***

6.380

(12.46)***

5.503

(10.78)***

8.176

(14.20)***

7.908

(13.95)***

5.849

(9.83)***

10.889

(40.57)***

SIZ

E0.290

(13.16)***

)0.109

()4.07)***

)0.896

()43.49)***

0.101

(4.66)***

)0.251

()10.04)***

)0.214

()8.61)***

)0.059

()2.25)**

)1.115

()95.31)***

r(CFO)

)9.995

()7.91)***

)1.558

()2.41)**

)5.592

()5.68)***

6.440

(8.54)***

0.540

(0.83)

)1.318**

()2.31)

)0.067

()0.09)

)0.073

()0.26)

r(SALES)

)2.365

()4.91)***

2.983

(5.28)***

4.297

(9.52)***

)3.830

()8.13)***

0.371

(0.69)

0.794

(1.45)

)0.097

()0.17)

)1.066

()4.56)***

OP-C

YCLE

)0.249

()2.98)***

0.166

(1.69)*

0.571

(6.57)***

0.124

(1.44)

)0.314

()3.22)***

)0.200

()2.13)**

)0.088

()0.87)

)0.068

()1.54)

NEG-EARN

)1.763***

()11.28)

)0.554***

()3.20)

)2.784***

()18.68)

)4.473***

()30.29)

)2.673***

()15.79)

)2.585***

()15.60)

)0.573***

()3.32)

0.302***

(3.98)

LEVERAGE

0.120

(1.64)

)0.297

()3.00)***

)0.474

()6.22)***

0.976

(10.60)***

)0.186

()2.04)**

0.013

(0.15)

)0.006

()0.07)

0.995

(15.66)***

MB

)0.008

()0.59)

0.022

(1.37)

0.106

(8.10)***

)0.015

()1.11)

)0.034

()2.23)**

)0.019

()1.26)

)0.016

()1.06)

)0.096

()6.70)***

INT-INTENSIT

Y0.178

(2.54)**

0.111

(2.41)**

0.244

(3.59)***

)0.030

()0.59)

0.071

(1.38)

0.100

(4.58)***

)0.048

()1.05)

)0.059

()2.22)**

INT-D

UMMY

0.586

(6.12)***

)0.063

()0.52)

0.301

(2.85)***

0.502

(4.88)***

0.213

(1.81)*

0.228

(1.98)**

)0.223

()1.96)*

0.127

(2.47)**

CAP-INTENSIT

Y2.622

(9.23)***

1.226

(3.55)***

0.489

(1.81)**

)1.575

()5.41)***

0.389

(1.17)

)0.566

()1.73)*

)0.655

()1.94)*

0.168

(1.16)

CONC

)2.236

()3.09)***

)1.702

()1.95)*

)2.276

()2.91)***

)6.042

()7.69)***

)2.799

()3.31)***

)1.931

()2.40)**

)1.981

()2.21)**

0.321

(0.86)

Yr.&Ind.

Dum

mies

R2

Included

0.366

Included

0.016

Included

0.316

Included

0.245

Included

0.080

Included

0.075

Included

0.010

Included

0.826

Refer

toTab

le2forvariab

ledefinitions.White’s(1980)

heteroscedasticity-consistentt-statistics

are

shownin

parentheses.*,

**an

d***representsign

ificance

atthe0.10,

0.05and0.01

levels,respectively.



Tab

le5

Resultsofcross-sectionalregressionofdecileranksofearnings

attributesonindustry

concentrationan

dhomogeneity

dummy

DependentVariable(N

=4989)

AQ

PERSIS-

TENCE

PREDIC

T-

ABIL

ITY

SMOOTH-

NESS

RELE-

VANCE

TIM

EL-

INESS

CONSER-

VATISM

C-SCORE

Intercept

6.279

(12.34)***

3.639

(6.01)***

6.496

(12.30)***

5.720

(10.87)***

8.633

(14.59)***

8.177

(14.11)***

5.625

(9.29)***

10.832

(39.35)***

SIZ

E0.296

(13.37)***

)0.115

()4.32)***

)0.894

()43.35)***

0.104

(4.81)***

)0.243

()9.73)***

)0.210

()8.42)***

)0.063

()2.37)**

)1.116

()95.14)***

r(C

FO)

)9.915

()7.87)***

)1.668

()2.61)**

)5.581

()5.65)***

6.492

(8.55)***

0.628

(0.98)

)1.261

()2.21)**

)0.122

()0.16)

)0.085

()0.30)

r(SALES)

)2.370

()4.92)***

3.015

(5.35)***

4.313

(9.54)***

)3.835

()8.13)***

0.382

(0.71)

0.794

(1.45)

)0.088

()0.15)

)1.067

()4.56)***

OP-C

YCLE

)0.273

()3.25)***

0.189

(1.93)**

0.563

(6.45)***

0.109

(1.27)

)0.346

()3.52)***

)0.218

()2.32)**

)0.074

()0.73)

0.064

()1.45)

NEG-EARN

)1.743

()11.19)***

)0.573

()3.32)***

)2.776

()18.60)***

)4.460

()30.15)***

)2.646

()15.62)***

)2.569

()15.49)***

)0.586

()3.39)***

0.298

(3.93)***

LEVERAGE

0.117

(1.59)

)0.290

()2.90)***

)0.473

()6.21)***

0.973

(10.55)***

)0.187

()2.07)**

0.011

(0.13)

)0.004

()0.04)

0.995

(15.65)***

MB

)0.008

()0.64)

0.023

(1.45)

0.106

(8.08)***

)0.016

()1.14)

)0.035

()2.23)**

)0.019

()1.28)

)0.016

()1.03)

)0.096

()6.69)***

INT-INTENSIT

Y0.175

(2.49)**

0.106

(2.34)**

0.239

(3.54)***

)0.031

()0.61)

0.063

(1.24)

0.096

(4.54)***

)0.048

()1.05)

)0.058

()2.20)**

INT-D

UMMY

0.590

(6.14)***

)0.101

()0.84)

0.280

(2.63)***

0.508

(4.91)***

0.194

(1.66)

0.225

(1.95)**

)0.232

()2.02)**

0.128

(2.49)***

CAP-INTENSIT

Y2.605

(9.08)***

1.041

(3.00)***

0.350

(1.28)

)1.565

()5.31)***

0.228

(0.68)

)0.615

()1.86)*

)0.684

()2.01)**

0.185

(1.25)

CONC

)5.450

()3.66)***

)2.831

()1.55)

)6.342

()3.87)***

)7.556

()5.06)***

)10.066

()6.72)***

)5.180

()3.24)***

)0.866

()0.52)

1.156

(1.64)

HOMO

)0.367

()3.26)***

0.419

(2.98)***

)0.104

()0.85)

)0.228

()1.84)*

)0.458

()3.42)***

)0.275

()2.06)**

0.237

(1.72)*

0.058

(0.99)

CONC*H

OMO

4.955

(2.91)***

0.113

(0.05)

5.198

(2.77)***

2.497

(1.41)

10.096

(5.56)***

4.723

(2.54)***

)2.044

()1.02)

)1.177

()1.40)

Yr.&Ind.

Dum

mies

R2

Included

0.367

Included

0.020

Included

0.318

Included

0.246

Included

0.085

Included

0.076

Included

0.011

Included

0.826

HOMO

=1when

HOMOGENEIT

Y>

samplemedian;0otherwise.Refer

toTab

le2forvariabledefinitions.White’s(1980)

heteroscedasticity-consistent

t-statistics

areshownin

parentheses.*,

**an

d***representsignificance

atthe0.10,0.05an

d0.01levels,respectively.



estimation of equation (2). We find that the coefficient on CONC is negative andsignificant for five of the eight earnings attributes. More strikingly, almost alltheir magnitudes (in absolute terms) are larger than those reported in Table 4,for example, AQ = )5.450 in Table 5 as compared to )2.236 in Table 4, whileTIMELINESS has a coefficient of )5.180 in Table 5 as compared to )1.931 inTable 4. These results are consistent with our conjecture that firms that operatein concentrated industries and compete in a heterogeneous environment enjoy ahigher level of monopolistic power and are associated with lower earningsquality.The estimated coefficients of the dummy variable HOMO in Table 5 are neg-

ative for most, but not all, of the attributes. In other words, in the hypotheticalcase where the industry concentration measure (CONC) is equal to 0, there isno consistent impact on earnings quality for firms competing in a homo-geneous environment. The coefficient on the interaction term CONC*HOMOis positive and significant for four of the eight earnings attributes. The resultssupport the idea that firms that operate in concentrated industries but competein a homogeneous environment (HOMO = 1) are associated with a higherlevel of competition than those competing in a heterogeneous environment(HOMO = 0) and that the earnings quality of the former group of firmsis higher than that of the latter. This evidence is supportive of ourHypotheses 1(b).

5. Additional and sensitivity tests

5.1 Information quality

So far, we have examined the impact of product market competition on vari-ous attributes of earnings quality. However, earnings (or accounting informationin general) constitute only a small portion of information investors use to assesstheir portfolio. In a recent paper, Ball and Shivakumar (2008) conclude that theaverage quarterly earnings announcement is associated with approximately 1 percent to 2 per cent of total annual information. They further point out that theincrease in the amount of information released during earnings-announcementwindows in recent years is due only in part to increased concurrent releases ofmanagement forecasts – a substantial amount of information is released in man-agement forecasts and in analysts’ forecast revisions prior to earnings announce-ments. Given that industry- and firm-specific information can be filtered to thepublic through different channels, and to the extent that product market compe-tition does have an impact on earnings quality, such impact should not be lim-ited to accounting information alone.The information environment of a firm is made up of both private and public

information. To examine the impact of product market competition on the qual-ity of information held by informed investors and analysts, we follow the ‘BKLSmodel’ introduced by Barron et al. (1998) and employ the precision of analysts’



public (ffiffiffi

hp

), private (ffiffi

sp

) and total (ffiffiffiffiffiffiffiffiffiffiffi

hþ sp

) information as proxies for informa-tion quality.18 These proxies are then regressed on our concentration and homo-geneity proxies together with SIZE, LEVERAGE and MB as control variables,19

and the results are summarized in Table 6. Coefficient estimates for CONC andthe interaction term CONC*HOMO are all consistent with those reported inTables 5 and 6. Irrespective of which proxy for information quality is used, thecoefficients for CONC are all negative and significant (

ffiffiffi

hp

= )2.442,ffiffi

sp

= )1.232 andffiffiffiffiffiffiffiffiffiffiffi

hþ sp

= )2.851) and have higher magnitudes (in terms ofabsolute value) when HOMO and CONC*HOMO are added to the regressions(ffiffiffi

hp

= )5.059,ffiffi

sp

= )4.309 andffiffiffiffiffiffiffiffiffiffiffi

hþ sp

= )7.669). These findings indicatethat a higher level of industry concentration is associated with lower public, pri-vate and total information quality. The coefficients for CONC*HOMO are allsignificantly positive (

ffiffiffi

hp

= 4.145,ffiffi

sp

= 4.545 andffiffiffiffiffiffiffiffiffiffiffi

hþ sp

= 7.365). Again,these findings are supportive of our earlier argument that firms from highly con-centrated and heterogeneous industries create an opaque public and privateinformation environment to avoid attention from competitors and political sanc-tions.

5.2 Sensitivity tests

As noted in Section 4.1, our sample is highly skewed in terms of firm size andis dominated by firms from three industries within the manufacturing sector(SIC codes 35, 36 and 38). As a robustness check, we replicate our tests on earn-ings quality by partitioning the sample into two groups based on whether (i) firmsize is above or below the sample median and (ii) the firm comes from a SICcode 35, 36 or 38 industry. Comparison of the unreported coefficient estimatesof the various variables between the two groups in each of the above sensitivitytests indicates that our findings are not qualitatively affected by firm size orindustry grouping, and the coefficient estimates of CONC and the interactionterm CONC*HOMO from each of the groupings are consistent with and exhibitthe same characteristics as those reported in Tables 4 and 5.As a final check on the robustness of our results, Table 2 indicates that the dis-

tributions of the earnings attributes are skewed and not normal. In particular,CONSERVATISM is very much affected by extreme values. To ensure that our

18 Refer to Barron et al. (2002) for the measurement and estimation of the various prox-ies. As the unreported estimated values of h and s are highly skewed, we follow Gu (2005)and use their square roots as information quality proxies. Descriptive statistics for theproxies are presented in Table 2. To control for outliers and the skewness of our sampledistribution, we run our regression tests using the decile ranks of the information qualityvariables, while actual values are used for the test and control variables.

19 Firm size, leverage, and market-to-book ratio are control variables commonly used inempirical tests of the BKLS model. Prior studies (Lys and Soo, 1995; Barth et al., 2001)show that these variables are related to the precision of analysts’ information.



Table

6

Resultsofcross-sectionalregressionofdecileranksofinform

ationqualityonindustry

concentrationandhomogeneity

dummy

DependentVariable(N

=5740)

ffiffiffi h

pffiffi s

pffiffiffiffiffiffiffiffiffiffiffi

hþs

pffiffiffi h

pffiffi sp

ffiffiffi h

pþs

Intercept

2.265

(9.49)***

2.027

(8.56)***

1.380

(6.05)***

2.159

(9.95)***

2.223

(8.85)***

1.765

(7.31)***

SIZ

E0.406

(18.94)***

0.471

(22.18)***

0.557

(27.18)***

0.412

(19.15)***

0.477

(22.36)***

0.567

(27.61)***

LEVERAGE

)0.430

()6.45)***

)0.292

()4.43)***

)0.556

()8.74)***

)0.424

()6.36)***

)0.290

()4.38)***

)0.549

()8.64)***

MB

0.011

(1.03)

)0.022

()2.02)**

0.029

(2.82)***

0.009

(0.83)

)0.024

()2.19)**

0.026

(2.48)**

CONC

)2.442

()3.22)***

)1.232

()1.64)*

)2.851

()3.93)***

)5.059

()3.54)***

)4.309

()3.04)***

)7.669

()5.63)***

HOMO

––

–)0.394

()3.08)***

)0.328

()2.59)***

)0.615

()5.05)***

CONC*H

OMO

––

–4.145

(2.45)**

4.545

(2.71)***

7.365

(4.57)***

Yr.&Ind.Dum

mies

R2

Included

0.089

Included

0.104

Included

0.168

Included

0.090

Included

0.105

Included

0.172

Refer

toTable

2forvariabledefinitions.HOMO

=1when

HOMOGENEIT

Y>

samplemedian;0otherwise.White’s(1980)

heteroscedasticity-consistent

t-statistics

areshownin

parentheses.*,

**an

d***representsign

ificance

atthe0.10,0.05and0.01levels,respectively.



regression results are not affected, we use decile ranks of earnings attributes inall our regression tests. Unreported regression results obtained using values ofearnings attributes instead of their decile ranks indicate that with the exceptionof CONSERVATISM, coefficient estimates for all earnings attributes are notqualitatively affected. Our overall interpretations and conclusions based on theseregression tests also remain unaffected.

6. Conclusion

The objective of this paper is to examine the impact of product market com-petition on earnings quality. Based on a sample from the US manufacturingsector for the period 1996–2005, we find consistent evidence showing a positiverelation between the level of product market competition and various attributesof earnings quality. We further provide evidence that earnings of firms compet-ing in concentrated and heterogeneous industries are associated with a lowerlevel of accruals quality, predictability, value relevance and timeliness than arethose of firms competing in concentrated and homogeneous industries. Thesefindings are consistent with the intuition that firms in concentrated and hetero-geneous industries tend to protect their competitive advantage by creating anopaque information environment. Sensitivity tests confirm that the above find-ings are robust to differences in firm size and industry grouping. Additional testsshow that competition not only helps to improve earnings quality but alsoassists in improving the precision of private and public information held byinvestors and analysts. Overall, the above evidence confirms that productmarket competition plays a major role in managers’ voluntary disclosure deci-sions and reinforces the idea that a manager’s motives for disclosure are influ-enced by the firm’s economic determinants and institutional settings, as well asby industry characteristics.One limitation of our study is that data used for the Herfindahl index

constructed for our sample were collected from Census of Manufacturespublications covering both private and public firms from the US manufacturingsector. This was done to ensure that the index more accurately reflects theactual market shares of firms within an industry. The extent to which ourempirical results can be generalized to other non-manufacturing sectors is anopen question.

References

Aggarwal, R. K., and A. A. Samwick, 1999, Executive compensation, strategic competi-tion, and relative performance evaluation: theory and evidence, Journal of Finance 54,1999–2043.

Aghion, P., N. Bloom, R. Blundell, R. Griffith, and P. Howitt, 2005, Competition andinnovation: an inverted-U relationship, The Quarterly Journal of Economics 120,701–728.



Akdogu, A., and P. Mackay, 2008, Investment and competition, Journal of Financial andQuantitative Analysis 43, 299–320.

Ali, A., S. Klasa, and E. Yeung, 2009, The limitations of industry concentration measuresconstructed with Compustat data: implications for finance research, Review of FinancialStudies 22, 3839–3871.

Ali, A., S. Klasa, and E. Yeung, 2010, Oligopolistic industries and corporate disclosurepolicy. [Cited 5 November 2010.] Available from URL: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1696844.

Bain, J. S., 1951, Relation of profit ratio to industry concentration: American manufac-turing 1936–1940, Quarterly Journal of Economics 65, 293–324.

Bain, J. S., 1968, Industrial Organization, 2nd edn (Wiley, New York).Ball, R., and L. Shivakumar, 2008, How much new information is there in earnings?,Journal of Accounting Research 46, 975–1016.

Barron, O. E., O. Kim, S. C. Lin, and D. E. Stevens, 1998, Using analysts’ forecasts tomeasure properties of analysts’ information environment, The Accounting Review 73,421–433.

Barron, O. E., D. Byard, and O. Kim, 2002, Changes in analysts’ information aroundearnings announcements, The Accounting Review 77, 821–846.

Barth, M. E., R. Kasznik, and M. McNichols, 2001, Analyst coverage and intangibleassets, Journal of Accounting Research 39, 1–34.

Basu, S., 1997, The conservatism principle and the asymmetric timeliness of earnings,Journal of Accounting and Economics 24, 3–37.

Botosan, C. A., and M. Stanford, 2005, Managers’ motives to withhold segment disclo-sures and the effect of SFAS No. 131 on analysts’ information environment, TheAccounting Review 80, 751–771.

Cahan, S. F., 1992, The effect of antitrust investigations on discretionary accruals: arefined test of the political-cost hypothesis, The Accounting Review 67, 77–95.

Campello, M., 2006, Debt financing: does it hurt or boost firm performance in productmarkets?, Journal of Financial Economics 82, 135–172.

Caves, R. E., 1980, Industrial organization, corporate strategy and structure, Journal ofEconomic Literature 18, 64–92.

Chan, L., J. Lakonishok, and T. Sougiannis, 2001, The stock market valuation ofresearch and development expenditures, Journal of Finance 56, 2431–2456.

Cohen, D. A., 2008, Does information risk really matter? An analysis of the determinantsand economic consequences of financial reporting quality, Asia-Pacific Journal ofAccounting & Economics 15, 69–90.

Core, J., W. Guay, and R. Verdi, 2008, Is accruals quality a priced risk factor? Journal ofAccounting and Economics 46, 2–22.

Darrough, M. N., and N. M. Stroughton, 1990, Financial disclosure policy in an entrygame, Journal of Accounting and Economics 12, 219–243.

Dechow, P., and I. Dichev, 2002, The quality of accruals and earnings: the role of accrualestimation errors, The Accounting Review 77(Suppl.), 35–59.

Dhaliwal, D. S., S. X. Huang, I. K. Khurana, and R. Pereira, 2008, Product market com-petition and accounting conservatism. [Cited 19 December 2008.] Available from URL:http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1266754.

Francis, J., R. LaFond, P. Olsson, and K. Schipper, 2004, Costs of equity and earningsattributes, The Accounting Review 79, 967–1010.

Francis, J., R. LaFond, P. Olsson, and K. Schipper, 2005, The market pricing of accrualsquality, Journal of Accounting and Economics 39, 295–327.

Francis, J., P. Olsson, and K. Schipper, 2006, Earnings quality, Foundations and Trends inAccounting 1, 259–340.



Gaio, C., and C. Raposo, 2011, Earnings quality and firm valuation: international evi-dence, Accounting & Finance 51, 467–499.

Garcia Lara, J. M., B. Garcia Osma, and A. Mora, 2005, The effect of earnings manage-ment on the asymmetric timeliness of earnings, Journal of Business Finance andAccounting 32, 691–726.

Gelb, D. S., and P. Zarowin, 2002, Corporate disclosure policy and the informativenessof stock prices, Review of Accounting Studies 7, 33–52.

Grossman, S. J., 1981, The role of warranties and private disclosure about product qual-ity, Journal of Law and Economics 24, 461–483.

Grossman, S. J., and O. D. Hart, 1980, Disclosure laws and takeover bids, Journal ofFinance 35, 323–334.

Gu, Z., 2005,Measuring the Precision of Analyst Private and Common Information: Gener-alization and an Application, Working paper (Carnegie Mellon University, Pittsburgh).

Hagerman, R. L., and M. E. Zmijewski, 1979, Some economic determinants of account-ing policy choice, Journal of Accounting and Economics 1, 141–161.

Hall, B., 1992, Investment and Research and Development at the Firm Level: Does theSource of Financing Matter? Working paper (National Bureau of Economic Research,Cambridge).

Han, J., and S. Wang, 1998, Political costs and earnings management of oil companiesduring the 1990 Persian Gulf crisis, The Accounting Review 73, 103–117.

Harris, M. S., 1998, The association between competition and managers’ business seg-ment reporting decisions, Journal of Accounting Research 36, 111–128.

Haushalter, D., S. Klasa, and W. Maxwell, 2007, The influence of product marketdynamics on a firm’s cash holdings and hedging behavior, Journal of Financial Econom-ics 84, 797–825.

Hay, D. A., and D. J. Morris, 1991, Industrial Economics and Organization, Theory andEvidence (Oxford University Press, Oxford).

Healy, P. M., and K. G. Palepu, 2001, Information asymmetry, corporate disclosure, andthe capital markets: a review of the empirical disclosure literature, Journal of Account-ing and Economics 31, 405–440.

Holmstrom, B., 1982, Moral hazard in teams, Bell Journal of Economics 13, 324–340.Hou, K., and D. T. Robinson, 2006, Industry concentration and average stock returns,Journal of Finance 61, 1927–1956.

Jacobsen, R., 1988, The persistence of abnormal returns, Strategic Management Journal9, 415–430.

Jehle, G. A., and P. J. Reny, 2000, Advanced Microeconomic Theory, 2nd edn (Addison-Wesley, Massachusetts).

Jones, J., 1991, Earnings management during import relief investigations, Journal ofAccounting Research 29, 193–228.

Khan, M., and R. Watts, 2009, Estimation and empirical properties of a firm-yearmeasure of accounting conservatism, Journal of Accounting and Economics 48, 132–150.

Kormendi, R., and R. Lipe, 1987, Earnings innovations, earnings persistence, and stockreturns, Journal of Business 60, 323–345.

LaFond, R., and R. Watts, 2008, The information role of conservative financial state-ments, The Accounting Review 83, 447–478.

Li, X., 2010, The impacts of product market competition on the quantity and quality ofvoluntary disclosures, Review of Accounting Studies 15, 663–711.

Lim, S., and Z. Matolcsy, 1999, Earnings management of firms subjected to product pricecontrols, Accounting & Finance 39, 131–150.

Lipczvnski, J., and J. Wilson, 2001, Industrial Organisation: An Analysis of CompetitiveMarkets (Pearson Education, Essex).



Lipe, R., 1990, The relation between stock returns and accounting earnings given alterna-tive information, The Accounting Review 65, 49–71.

Lys, T., and L. G. Soo, 1995, Analysts’ forecast precision as a response to competition,Journal of Accounting, Auditing and Finance 10, 751–765.

MacKay, P., and G. M. Phillips, 2005, How does industry affect firm financial structure?Review of Financial Studies 18, 1433–1466.

Marquardt, C. A., and C. I. Wiedman, 2004, The effect of earnings management on thevalue relevance of accounting information, Journal of Business Finance and Accounting31, 297–332.

Milgrom, P., 1981, Good news and bad news: representation theorems and applications,Bell Journal of Economics 12, 380–391.

Miller, M., and K. Rock, 1985, Dividend policy under asymmetric information, Journalof Finance 40, 1031–1051.

Nalebuff, B. J., and J. E. Stiglitz, 1983, Prizes and incentives: towards a general theory ofcompensation and competition, Bell Journal of Economics 14, 21–43.

Parrino, R., 1997, CEO turnover and outside succession: a cross-sectional analysis, Jour-nal of Financial Economics 46, 165–197.

Penman, S. H., 2001, Financial Statement Analysis and Security Valuation (McGraw-Hill/Irwin, New York).

Porter, M. E., 1979, The structure within industries and companies’ performance, Reviewof Economics & Statistics 61, 214–227.

Qualls, D., 1974, Stability and persistence of economic profit margins in highly concen-trated industries, Southern Economic Journal 40, 604–612.

Roberts, P. W., and G. R. Dowling, 2002, Corporate reputation and sustained superiorfinancial performance, Strategic Management Journal 23, 1077–1093.

Scherer, F. M., 1980, Industrial Market Structure and Economic Performance, 2nd edn(Houghton Mifflin, Boston).

Schumpeter, J. A., 1934, The Theory of Economic Development (Harvard University Press,Cambridge).

Subramanyam, K. R., 1996, The pricing of discretionary accruals, Journal of Accountingand Economics 22, 249–281.

Trueman, B., and S. Titman, 1988, An explanation for accounting income smoothing,Journal of Accounting Research 26(Suppl.), 127–139.

Verrecchia, R., 1983, Discretionary disclosure, Journal of Accounting and Economics 5,179–194.

Warfield, T., J. J. Wild, and K. J. Wild, 1995, Managerial ownership and the informative-ness of earnings, Journal of Accounting and Economics 20, 61–92.

Watts, R., 2003, Conservatism in accounting, part 1: explanations and implications,Accounting Horizons 17, 207–221.

Watts, R., and J. Zimmerman, 1978, Towards a positive theory of the determination ofaccounting standards, The Accounting Review 53, 112–134.

Weiss, L. W., 1974, The Concentration-Profits Relationship and Antitrust, in: H. Golds-chmid, H. Mann, H. Weston, eds, Industrial Concentration: The New Learning (LittleBrown, Boston), 184–233.

Weiss, L. W., 1979, The structure-conduct-performance paradigm and antitrust, Univer-sity of Pennsylvania Law Review 127, 1104–1140.

White, H., 1980, A heteroskedasticity-consistent covariance matrix estimator and a directtest for heteroskedasticity, Econometrica 48, 817–838.

